Glutamate Transporters and Mitochondria: Signaling, Co-compartmentalization, Functional Coupling, and Future Directions

Abstract

In addition to being an amino acid that is incorporated into proteins, glutamate is the most abundant neurotransmitter in the mammalian CNS, the precursor for the inhibitory neurotransmitter γ-aminobutyric acid, and one metabolic step from the tricarboxylic acid cycle intermediate α-ketoglutarate. Extracellular glutamate is cleared by a family of Na⁺-dependent transporters. These transporters are variably expressed by all cell types in the nervous system, but the bulk of clearance is into astrocytes. GLT-1 and GLAST (also called EAAT2 and EAAT1) mediate this activity and are extremely abundant proteins with their expression enriched in fine astrocyte processes. In this review, we will focus on three topics related to these astrocytic glutamate transporters. First, these transporters co-transport three Na⁺ ions and a H⁺ with each molecule of glutamate and counter-transport one K⁺; they are also coupled to a Cl⁻ conductance. The movement of Na⁺ is sufficient to cause profound astrocytic depolarization, and the movement of H⁺ is linked to astrocytic acidification. In addition, the movement of Na⁺ can trigger the activation of Na⁺ co-transporters (e.g. Na⁺ - Ca²⁺ exchangers). We will describe the ways in which these ionic movements have been linked as signals to brain function and/or metabolism. Second, these transporters co-compartmentalize with mitochondria, potentially providing a mechanism to supply glutamate to mitochondria as a source of fuel for the brain. We will provide an overview of the proteins involved, discuss the evidence that glutamate is oxidized, and then highlight some of the un-resolved issues related to glutamate oxidation. Finally, we will review evidence that ischemic insults (stroke or oxygen/glucose deprivation) cause changes in these astrocytic mitochondria and discuss the ways in which these changes have been linked to glutamate transport, glutamate transport-dependent signaling, and altered glutamate metabolism. We conclude with a broader summary of some of the unresolved issues.

Introduction

It is well-established that the amino acid glutamate is the predominant excitatory neurotransmitter in the mammalian CNS, while γ-aminobutyric acid (GABA) is the predominant inhibitory neurotransmitter in brain (1–3). There are several remarkable aspects of these neurotransmitters. First, the levels of glutamate and GABA are extraordinarily high. Glutamate levels are 5–10 mmol/Kg and the levels of GABA are approximately 1 mmol/Kg (4). If the brain were 100% water and glutamate were uniformly distributed, the levels of glutamate would be 10 mM. This is 1,000- to 10,000-fold higher than those of other important neurotransmitters, including all of the biogenic amines. It is worthwhile to note that the levels of glucose, the accepted predominant fuel for brain, are only ~1 mM in the brain (5). Second, glutamate is only one metabolic step from GABA. Third, both these neurotransmitters can be converted to fuel that can be oxidized in mitochondria. In fact, many of the enzymes that metabolize glutamate are found in mitochondria, and the enzyme that metabolizes GABA, GABA-transaminase, is a mitochondrial enzyme, converting GABA to succinyl semialdehyde while also converting α-ketoglutarate to glutamate. Finally, it is clear that an extracellular accumulation of glutamate is sufficient to cause neuronal cell death through a mechanism known as ‘excitotoxicity’ (6–10).

Although several studies have detected micromolar extracellular concentrations of glutamate using microdialysis probes (11, 12), analyses of baseline glutamate receptor activation put the number at closer to 25 nM (13–15). With total brain concentrations of 10 mmol/kg, this means intracellular concentrations of glutamate are ~250,000-fold higher than those in the extracellular space. This is an oversimplification because glutamate is packaged into vesicles by vesicular glutamate transporters and glutamate is likely quickly metabolized in astrocytes (16). There is still no evidence for extracellular metabolism of glutamate (2); therefore, all released glutamate needs to be cleared from the extracellular space by transporters. There are two families of these transporters, the first are Na⁺-symporters and the second exchange glutamate for cystine. These transporters have been the subject of several different reviews (17–27). We will focus on the family of Na⁺-dependent transporters. In brief, these transporters are called GLAST, GLT-1, EAAC1 (alternatively, these are called EAAT1, EAAT2, and EAAT3, respectively), EAAT4, and EAAT5. Some investigators use the two nomenclatures to selectively refer to rodent or human homologs, respectively, but this is not always the case. Both GLT-1 and GLAST are enriched in glial cells in the nervous system, but lower levels of GLT-1 are also found on neurons (28, 29). Interestingly, GLT-1 requires only 100-fold enrichment to obtain purified protein; this indicates that GLT-1 represents 1% of brain protein (30). The energetic cost of maintaining very low levels of extracellular glutamate is quite high, and this is accomplished by capturing the energy from moving three molecules of Na⁺ down its concentration gradient with each molecule of glutamate. In addition, the transport of glutamate is accompanied by the inward movement of one H⁺ and the counter-transport a K⁺ with each cycle (31).

Given the high level of transport activity, it should not be surprising that the transport of glutamate into astrocytes has significant implications for astrocytic ionic signaling and astrocytic metabolism (Fig. 1). These two topics will be first addressed below. This will be followed by a summary of relatively recent observations that suggest that glutamate transport triggers astrocytic mitochondrial fragmentation after hypoxia/ischemia.

Fig. 1.

Schematic representation of the effects of glutamate uptake into astrocytes. Glutamate (Glu) uptake into astrocytes through Na⁺-dependent transporters is accompanied by an influx of 3 molecules of Na⁺ and one H⁺. The concentration gradient of Na⁺, the driving force behind Glu uptake, is established by the Na⁺/K⁺ ATPase. Increased intracellular Na⁺ causes reversed operation of the Na⁺/Ca²⁺ exchanger (NCX), resulting in increased intracellular Ca²⁺, which in turn can be taken up into mitochondria, increase glucose transport into the astrocyte, and/or cause vesicular release of ATP or gliotransmitters (including Glu) out of the astrocyte. These Glu transporters can also conduct Cl⁻, and this is especially relevant in the neuronal transporters where it can inhibit depolarization at the presynaptic nerve terminal. Upon acute injury such as ischemia, there is increased extracellular Glu and increased transport, which results in changes in astrocytic mitochondrial morphology and either degradation of the mitochondria through mitophagy or vesicular release for transfer to other cells, including neurons.

Glutamate transporters as signaling molecules

After release, glutamate diffuses out of the synapse and/or is cleared by the Na⁺-dependent transporters (for review, see 32). When one considers glutamate-mediated, inter-cellular signaling, one generally thinks of ligand-gated ion channels and G-protein coupled, so-called ‘metabotropic’ receptors (for reviews, see 33–39). In addition to limiting activation of these prototypic glutamate receptors, there is now substantial evidence that activation of transport, itself, causes downstream signaling. Transporter cycle times are relatively slow; even the fastest require between 10 and 75 msec to move a molecule of the glutamate across the plasma membrane (40, 41). However, the levels of transporters are high enough that the clearance of either exogenous or endogenous glutamate results in significant astrocytic depolarization (41–43). The membrane is repolarized by K⁺ channels and by the Na⁺/K⁺ ATPase (44), which has the stoichiometry of 3 Na⁺ (outward) and 2 K⁺ (inward) for each molecule of ATP hydrolyzed (45). This movement of Na⁺ and the depolarization both places a substantial energetic demand on the cell and contributes to signaling within and between astrocytes.

Exogenous glutamate or the transportable glutamate transporter substrate D-aspartate cause increases in intracellular Na⁺ in primary astrocyte cultures (46), as well as in astrocytes in acute hippocampal slices (47). Neuronal activity also causes an increase in intracellular Na⁺ in hippocampal slices. This effect is almost completely blocked by the non-transportable blocker of glutamate transporters threo-β-benzyloxyaspartic acid (TBOA). Together, these results indicate that neuronal release of endogenous pools of glutamate activate Na⁺ signals in astrocytes. These signals rapidly propagate in processes at speeds of up to 120μm/sec (48 for review, see 49), and the “spread” of the Na⁺ signal is proportional to the intensity of the stimulation (47). Interestingly, Ziemens et al. recently showed that the increases in Na⁺ in cortical astrocytes were approximately twice the size of those observed in hippocampal astrocytes. Furthermore, they showed that inhibition of glutamate transport activity only partially blocked (~50%) the signal in cortical astrocytes; the remaining 50% was blocked by an NMDA receptor antagonist (50). This suggests that some subtypes of astrocytes differentially engage different sources of glutamate-evoked Na⁺ signaling. Finally, the short-term plasticity of these transport-dependent Na⁺ currents are sensitive to the intracellular Na⁺ concentrations, suggesting that other mechanisms of Na⁺ transport could modulate glutamate transport and the accompanying Na⁺ influx (51).

While it is well-established that this influx of Na⁺ serves to drive the transport of glutamate against its concentration gradient, an energetically-demanding process, there is also evidence that these Na⁺ currents serve additional signaling purposes within the astrocyte, particularly in metabolic signaling. Bernardinelli et al. showed that glutamate transport-dependent increases in astrocytic Na⁺ are associated with increased Ca²⁺ and increased glucose uptake (52). They suggest a model in which increases in intracellular Ca²⁺ cause the release of ATP and glutamate, followed by glutamate uptake (along with Na⁺) and ATP-signaling, thereby increasing intracellular Ca²⁺ and propagating the signal from astrocyte to astrocyte in a so-called ‘metabolic wave.’ Porras et al. also observed a glutamate- or D-aspartate-dependent increase in astrocytic glucose transporter (GLUT1) activity. This effect is blocked by an inhibitor of glutamate transport and is mimicked by increasing both Na⁺ and Ca²⁺ simultaneously in astrocytes (53). Several groups have documented a glutamate transport-dependent increase in glycolysis resulting in release of lactate, both in astrocyte cultures and brain slice preparations (54–56). This is the so-called ‘astrocyte-neuron lactate shuttle’. The fact that this is not consistently observed (57–61) suggests that an uncontrolled variable regulates the extent to which glutamate transport couples to glycolysis. Glutamate uptake also triggers an increase in NADH/NAD⁺ in astrocytes (62). This effect is dependent upon activation of the Na⁺-HCO₃⁻-cotransporter NBCe1.

While glutamate transporters are not directly coupled to the movement of Ca²⁺, the glutamate transport-dependent influx of Na⁺ activates the reverse operation of the Na⁺/Ca²⁺ exchanger (NCX), whereby extracellular Ca²⁺ moves into the astrocyte and Na⁺ is extruded (63, 64 for reviews, see 65, 66). Ibáñez and colleagues observed that the glutamate transport-dependent increase in Ca²⁺ is only partially (~50%) blocked by an inhibitor of reversed operation of NCX in astrocytes; the remainder is dependent upon other sources (67). This implies that glutamate transport is also indirectly linked to activation of other pathways that increase astrocytic Ca²⁺; these are not well defined at this time. This group also showed that the inhibitor of reversed operation of NCX blocked activity-dependent internalization of GLT-1. In addition to regulation of glutamate transport, it is clear that Ca²⁺ plays several additional roles in astrocyte signaling. The glutamate transport-induced increase in Ca²⁺ through reversed operation of NCX also contributes to the maturation of oligodendrocytes (68). Ca²⁺ signals have been shown to trigger release of gliotransmitters, including glutamate, which can in turn cause an increase in Ca²⁺ levels in neurons (Fig. 1) (69, 70); these astrocytic Ca²⁺ signals have also been linked to modulation of synaptic activity (71). As previously mentioned, Ca²⁺ oscillations have been linked to release of ATP from astrocytes, leading to propagation of the Ca²⁺ waves by inducing Ca²⁺ release from the endoplasmic reticulum (72). Finally, astrocytes serve as the intermediary between neurons and blood vessels in neurovascular coupling, and astrocytic Ca²⁺ contributes to this process (73), for reviews, see 74), 75).

The ionic signals discussed above are not limited to the cytosol of astrocytes. In fact, there is evidence that these signals have effects on mitochondria. As will be discussed in the next section of this review, there is a close physical association between glutamate transporters, the Na⁺/K⁺ ATPase (which is responsible for establishing the Na⁺ gradient necessary for glutamate transport), and mitochondria (76, 77). Bernardinelli et al. found that Na⁺ signals propagate to mitochondria in a Ca²⁺-dependent manner (78). The same group later observed spontaneous Na⁺ spiking activity in mitochondria that was actually decreased by glutamate, unaffected by TBOA, and positively correlated with increases in ATP (79). We showed that glutamate transport and the reversed operation of NCX decrease mitochondrial mobility (80). These changes in ionic concentrations almost certainly also regulate mitochondrial metabolism, but this is still relatively unexplored.

Finally, several members of the EAAT family also function as glutamate-activated Cl⁻ channels. In fact, the cycle times for one of the neuronal transporters is so slow that it seems unlikely that these transporters contribute much to the clearance of glutamate; rather, it is more likely that they function as ligand-gated anion channels to reduce excitability (81). EAAT4 is localized to Purkinje cells in the cerebellum and conducts Cl⁻ in a manner that is activated by neuronal activity (82–85). Similarly, EAAT5 expression is essentially restricted to the presynaptic terminal of bipolar cells and photoreceptors in the retina (86). It has been observed that glutamate-generated Cl⁻ currents in photoreceptors hyperpolarize the cells from which they originate (87), and that Cl⁻ conductance through EAAT5 in bipolar cells serves to oppose depolarization and inhibit synaptic transmission (88, 89), indicating a possible inhibitory function of EAAT5 in these cells, apart from merely taking up glutamate. The neuronal EAATs are not the only members of the family that exhibit anion fluxes—all five of the EAATs display an anion conductance (90, 91). In fact, Untiet et al. demonstrated that the astrocytic transporters EAAT1/GLAST and EAAT2/GLT-1 are important for Cl⁻ balance and resting membrane potential in developing Bergmann glia (92).

Glutamate transporters and mitochondria co-compartmentalize

The levels of GLT-1 and GLAST (EAAT2 and EAAT1, respectively) mRNA and protein are consistently higher in astrocytes than in other cells in the nervous system (for reviews, see 17, 18, 29, 93, 94). Furthermore, electron microscopic studies have documented that expression of both these transporters is enriched in fine processes of astrocytes (95, 96). Approximately a decade ago, our laboratory used an anti-GLT1 antibody to immunoprecipitate GLT-1 from rat cortex and identify proteins that coimmunoprecipitate with GLT-1 by mass spectrometry. We had hypothesized that GLT-1 might form complexes with signaling molecules previously implicated in regulation of GLT-1 activity (e.g. protein kinase C), proteins that formed scaffolds for GLT-1, and possibly proteins that support the ion gradients needed to fuel the activity of this transporter (e.g. the Na⁺/K⁺ ATPase). While some of the subunits of the Na⁺/K⁺ ATPase were identified, we unexpectedly identified many mitochondrial proteins and most of the enzymes in glycolysis (77). Around this same time, Erin Rose from Dave Hampson’s group performed a mass spectroscopy analysis of proteins that co-purified with the Na⁺/K⁺ ATPase from cerebellum and identified GLAST and many other proteins that overlapped with our analysis (76). Several things prompted us to doubt our results. First, our colleagues/collaborators had a hard time believing that mitochondria might form a complex with glutamate transporters. Although there is some precedent for mitochondrial interactions with a subtype of glucose transporter (97, 98), this was, and remains, a reasonable concern because no direct linkages have documented to date. Second, as described above, several observations had prompted the hypothesis that astrocytes might preferentially metabolize glucose to lactate and export the lactate for oxidative phosphorylation (99, 100). Finally, although mitochondria had been identified in fine astrocyte processes by electron microscopy (101–103), some investigators made the reasonable assumption that these very fine processes were too small to accommodate mitochondria (104). At about the time that we had confirmed many of the interactions with reverse immunoprecipitations, Dr. Joshua Jackson joined the laboratory. We spent considerable time trying different sets of experiments (trying to map domains of GLT-1 required for the various interactions, studying lateral mobility of GLT-1 in the plasma membrane, etc). Ultimately, Dr. Jackson prepared organotypic slice cultures and selectively transduced astrocytes with fluorescently tagged molecules. This led to several observations. First, ~45% of the typical astrocyte process is occupied by mitochondria. Second, ~42% of the typical process is occupied by fluorescently-tagged GLT-1. By definition, this means there will be considerable overlap of these two signals by chance. In fact, we found that GLT-1 puncta overlap mitochondria ~60% of the time, and using Monte Carlo simulations, we found that this overlap was significantly more than would occur by chance (77). Similarly, we found that mitochondria also overlap with exogenously expressed/fluorescently tagged GLAST (105). While this does suggest that there is significant co-compartmentalization of mitochondria with GLT-1 and GLAST under these circumstances, there are several issues that have not been addressed. First, these experiments were conducted with exogenously expressed GLT-1 or GLAST. Second, these organotypic slices have many advantages, but they are grown under conditions that are quite different from those observed in vivo (e.g. concentration of glucose, concentration of oxygen, lack of blood vessels, etc). Finally, it is still not clear if the co-immunoprecipitation is an artifact of solubilization followed by aggregation or if it reflects a true interaction.

The logical next direction for this work has been to test for the presence of mitochondria in fine astrocyte processes in vivo. Using a couple of different strategies, several groups have confirmed that mitochondria are found throughout astrocyte processes in vitro and in vivo. A few of us have found that increasing neuronal activity recruits mitochondria to locations in astrocyte processes that are adjacent to synapses (80, 106–113). A recent study also demonstrated that deletion of neuronal pools of GLT-1 results in a doubling of the mitochondrial occupancy of astrocytic processes near excitatory synapses (114). It is not clear if this effect is due to increased neuronal activity or is dependent upon some other mechanism. To date, the extent of overlap of endogenous GLT-1 or GLAST with mitochondria has not been examined in vivo and given the high levels of GLT-1 and GLAST that are found in astrocytes, this may not be possible. Nevertheless, these studies provide strong evidence that astrocytic glutamate transporters will be overlapped by (co-compartmentalize with) mitochondria.

Implications of co-compartmentalization for glutamate metabolism

The glutamate-glutamine cycle was identified as the pathway for recycling of neurotransmitter pools of glutamate (115, 116). After clearance into astrocytes, glutamate is converted to glutamine by glutamine synthetase. The glutamine is then exported to neurons and re-converted to glutamate (117). This pathway is still the best explanation for the current data and likely represents a major, or the only, pathway for maintaining neurotransmitter pools of glutamate. Remarkably, there are still some issues with this pathway. First, it is relatively hard to demonstrate that excitatory, glutamate-mediated, transmission is dependent upon glutamine (118, 119). In addition, the glutamine transporters that were cloned in the early 2000’s are found on the cell body and are not enriched in presynaptic terminals; this makes it hard to believe that they contribute to recycling of neurotransmitter pools of glutamate (117, 120). A recent study from Jeff Erickson seems to be heading the field toward an activity that could mediate presynaptic glutamine uptake (121).

In addition to glutamine synthetase, there are several different enzymes and transporters involved in glutamate metabolism in astrocytes (Fig. 2). Glutamate can be converted to α-ketoglutarate in the cytosol by aspartate aminotransferase (also called glutamate-oxaloacetic acid transaminase); this is a pyridoxal phosphate (B6/thiamine)-dependent enzyme that simply transfers the amine to oxaloacetate to generate aspartate. These enzymes are readily reversible and the direction of flux is only driven by concentrations of substrates and products. The resulting cytosolic α-ketoglutarate can be imported by a mitochondrial transporter (mitochondrial 2-oxoglutarate/malate carrier protein, M2OM, or Slc25a11) (122). Glutamate can also be directly imported into mitochondria. There are two types of mitochondrial glutamate transporters. One type co-transports a H⁺ and glutamate into mitochondria (glutamate carrier 1, GC1, or Slc25a22). The second exchanges glutamate and aspartate (aralar or Slc25a12) across the mitochondrial membrane. There are several studies that suggest that aralar is not expressed by astrocytes, but there are also several studies that observe either protein or mRNA for aralar in astrocytes (94, 123–125). There are two mitochondrial enzymes that can convert glutamate to α-ketoglutarate (126). The first is a mitochondrial version of aspartate aminotransferase and the second is glutamate dehydrogenase. Unlike aspartate aminotransferase, the conversion of glutamate to α-ketoglutarate by glutamate dehydrogenase is associated with generation of NAD(P)H which can be converted to ATP or used as reducing power. The reader is referred to the following excellent discussions of the contributions of these enzymes to glutamate oxidation (126–128). It seems likely that both enzymes contribute to glutamate oxidation but under different, not well-defined conditions.

Fig. 2.

Schematic illustration of glutamate metabolism in astrocytes. In the cytosol, glutamate (Glu) can be transformed into glutamine (Gln) by glutamine synthetase (GS). This Gln is released into the extracellular space for subsequent transport into neurons and replenishment of neuronal Glu. Glutamate can also be transformed into α-ketoglutarate (αKG) by cytosolic aspartate aminotransferase (AST) which is transported into mitochondria via the mitochondrial 2-oxoglutarate/malate carrier protein (M2OM/Slc25a11). Glu can also be transported into mitochondria directly, either by glutamate carrier 1 (GC1) or by the glutamate/aspartate exchanger (aralar). In mitochondria, Glu is transformed into αKG by AST or by glutamate dehydrogenase (GDH). αKG is converted to malate which is exported into the cytosol and converted to pyruvate via malic enzyme (MEc) that can either be converted to lactate or to acetyl CoA by pyruvate dehydrogenase complex (PD) for complete oxidation. Glu (five carbon backbone) is replenished by converting pyruvate to oxaloacetate (OAA) by the astrocyte specific enzyme pyruvate carboxylase (PC) and the OAA is converted to αKG. The αKG can be converted to glutamate by either AST or GDH.

The α-ketoglutarate that is generated from glutamate is either partially oxidized to yield 5–7 molecules of ATP or completely oxidized to ultimately generate 20 molecules of ATP (129, 130). This occurs because tricarboxylic acid cycle intermediates cannot be depleted or accumulated. So after α-ketoglutarate is partially oxidized to malate, the malate needs to exit the mitochondria to the cytosol where it is converted to pyruvate (129). This pyruvate can either be converted to acetyl-CoA for re-entry into tricarboxylic acid cycle by pyruvate dehydrogenase (PD) or converted to lactate which could be exported into the extracellular space (131, 132).

The co-compartmentalization of glutamate transporters with mitochondria would seemingly provide an ideal mechanism for glutamate to be used as fuel to support its own uptake and provide ATP for glutamine synthesis. In fact, there are several different observations that demonstrate that some percentage of glutamate enters mitochondria and is oxidized. First, glutamate oxidation has been directly measured in astrocyte cultures either using glutamate that is radioactively labeled on the 1 carbon carboxyl group; one can capture and quantify the CO₂ that is generated. Alternatively, non-radioactive, labeled glutamate can be used to follow glutamate oxidation using either NMR or mass spectrometry (133, 134 for reviews, see 104, 135). The percentage of glutamate that is oxidized is somewhat variable between laboratories ranging between 10% and 60% of the glutamate that enters an astrocyte (for review and primary references, see 75). Interestingly the percentage of glutamate oxidation increases with increasing extracellular glutamate, consistent with the notion that the amount of glutamate that is oxidized can be regulated (136, 137). Second, deletion of glutamate carrier 1 (Slc25a22) results in intracellular accumulation of glutamate in astrocytes, suggesting that this transporter is used to provide glutamate to mitochondria for oxidation (138). Finally, there has to be a net preservation of glutamate in the brain and it is generally thought that glutamate is not transported into or out of the brain, so one can estimate the amount of glutamate oxidation by following the incorporation of isotope from peripherally administered glucose into glutamate in vivo (139). The glucose supplies pyruvate to astrocytes where it can be converted to oxaloacetate by the astrocytic enzyme, pyruvate carboxylase, this is then conjugated with acetyl-CoA to generate α-ketoglutarate which can then be converted to glutamate (140, 141). This is the pathway that replenishes neuronal glutamate. Flux through this pathway has also been used to estimate that about 10 to 30% of brain glutamate is oxidized on a continuous basis (for reviews, see 115, 129, 130, 142).

While there are differences in both the results and conclusions with the studies of glutamate metabolism, it seems likely that the differences are related to differences in the conditions under which the results were obtained. For example, many of the critical studies were performed in astrocyte-enriched cultures. From a metabolic perspective, this is the only system that allows specific interrogation of ‘astrocyte’ metabolism in the absence of contributions from other cells. While it seems clear that this will remain an important model system into the future, individuals who use this system will need to be aware of several issues. First, from a mRNA expression perspective, cultured astrocytes are quite different from those observed in vivo (143). We and several other groups have demonstrated that both neurons and endothelia affect expression of various proteins in astrocytes (144–146). In fact, co-culturing astrocytes with neurons increases astrocytic expression of several glycolytic enzymes (147). Second, the conditions under which these experiments are conducted can have consequences for the results. For example, the levels of O₂ are known to change astrocyte Ca²⁺ signaling (148) and Ca²⁺ can activate several of the mitochondrial proteins involved in glutamate metabolism, including aralar and several of the mitochondrial dehydrogenases (149–154 for recent discussion, see 113). Although changing glucose concentrations does not affect the extent of glutamate oxidation in astrocyte cultures (155), it seems likely that glucose levels could easily affect measures of energy production. Finally, the buffers and ionic composition of the medium can also affect the results as illustrated by Rimmele and her colleagues (156). One suspects that as genetically-encoded indicators of brain metabolites are optimized, it will be possible to determine exactly how endogenous pools of glutamate affect astrocytic oxidative phosphorylation, to monitor lactate production from endogenous glutamate as opposed to endogenous glucose, and to test for possible linkages between neuronal activation and the relative disposition of glutamate.

While it is still not clear that co-compartmentalization of glutamate transporters with mitochondria necessarily favors import of glutamate into mitochondria and subsequent oxidation, there is evidence that glutamate transport is coupled to mitochondrial function. We found that several different inhibitors of glutamate dehydrogenase non-competitively block glutamate or D-aspartate uptake, consistent with the notion that endogenous pools of glutamate are oxidized to provide fuel for glutamate uptake, at least under some conditions (157). Conversely, glutamate transport also changes mitochondrial function. Originally, glutamate transport activation was linked to mitochondrial acidification, decreased oxygen consumption, and decreased ATP production (158). The mitochondrial import of H⁺ is associated with decreased oxidative phosphorylation by a mechanism that involves uncoupling protein 4 (159). These investigators also showed that D-aspartate, a substrate for the plasma membrane transporters but not for the mitochondrial transporters, also causes mitochondrial acidification. These studies were performed in a HEPES-based buffer. More recently some of these same investigators examined the effects of glutamate uptake on mitochondrial function in bicarbonate-based buffer. Under these conditions, glutamate triggers increased mitochondrial function as assessed by measuring the oxygen consumption rate (156). A similar glutamate transport-dependent increase in mitochondrial function (oxygen consumption) was recently observed by Juaristi and colleagues (132). They also showed that D-aspartate triggers an increase in oxygen consumption, consistent with the notion that this increase is not dependent upon glutamate per se but is instead triggered by the energetic demands imposed by the increase in Na⁺/K⁺ ATPase activity required to maintain Na⁺ gradients in the cell (for review, see 49).

While one might intuitively assume that the coupling of glutamate transport with mitochondrial function and the potential to oxidize glutamate (see above) would result in glutamate transport-dependent increases in ATP, this is not the case. In fact, glutamate transport activation consistently causes a decrease in astrocytic ATP, regardless of whether D-aspartate or L-glutamate is used as a transporter substrate (47, 132, 160, 161). In addition, neuronal activity and the Na⁺/K⁺ ATPase both increase extracellular K⁺. Increased extracellular K⁺ rapidly stimulates glycolytic activity and increases astrocytic ATP (54, for reviews see 162 163, 164, 165); it seems likely that this glycolytic activity is directly fueling transport. Together, these data suggest that the amount of glutamate that enters via the transporter and is oxidized does not generate excess ATP beyond that required to support re-generation of the Na⁺ gradient through the Na⁺/K⁺ ATPase and fuel the conversion of glutamate to glutamine. All of these studies were conducted either in cultured astrocytes or acute brain slices; it is possible that the balance of ATP generated changes in an in vivo setting, but this has not been tested to date. The effects of glutamate and/or K⁺ on ATP levels may follow different kinetics. In non-astrocyte systems, Magi and colleagues have demonstrated the presence of EAAC1 in purified mitochondria, and have documented a glutamate transport- and NCX-dependent increase in mitochondrial Ca²⁺ and ATP. They also show that EAAC1 co-immunoprecipitates with NCX1 (166–168). Together, these findings suggest that glutamate transport is physically and functionally coupled to Ca²⁺ influx via NCX1, and that activation of this system can trigger an increase in ATP in some cells.

Effects of Hypoxia/ischemia on astrocytic glutamate transport and mitochondria

The brain is highly dependent upon energy. In fact, it consumes ~20% of the basal metabolic rate, and it is generally thought that it is almost exclusively dependent upon glucose for energy production. (169–172). Remarkably, the brain contains relatively small reserve pools of glycogen, and these pools are essentially restricted to astrocytes (173, 174). Therefore, it should not be surprising that the brain is particularly sensitive to periods of hypoxia and/or decreased blood flow (ischemia). Over 30 years ago, several groups simultaneously demonstrated that glutamate receptor antagonists protect neurons from global and/or focal ischemia in in vitro and in vivo model systems (for reviews, see 6, 7, 175, 176). While this and several other lines of evidence are consistent with the notion that these insults cause a rise in extracellular concentrations of glutamate and excessive activation of glutamate receptors followed by excitotoxic cell death, excitatory amino acid receptor antagonists have not made it into clinical practice for reasons that are still unclear (177, 178).

Compared to neurons, astrocytes are relatively resistant to periods of hypoxia/ischemia. This relative sparing has been observed in vitro and in vivo (179, for reviews see 180, 181, 182). Depending on their location relative to a focal ischemic infarct, astrocytes undergo a progressive change in morphology with loss of highly branched processes, hypertrophy, and increased expression of glial fibrillary acidic protein (GFAP) (183 for review, see 184). These are termed ‘reactive astrocytes’. Astrocytes also proliferate and can generate new neurons (185–187 for review, see 188). Several different groups have observed decreases in the levels of GLT-1 and/or GLAST mRNA and/or protein in several different models of hypoxia/ischemia (189, 190 191 for reviews, see 22, 192). It is interesting to note that decreases in protein frequently occur within 24 hours after an insult. Although not necessarily translatable to in vivo, the half-lives of GLT-1 and GLAST proteins are longer than 24 h when measured in astrocyte cultures (193); this suggests that these insults may accelerate degradation of these transporters possibly by ubiquitination in addition to the decreased translation that presumably accompanies the decreases in mRNA for these transporters (194–196).

The effects of hypoxia and ischemia on astrocytic mitochondria have also been examined in vitro and in vivo. In primary cultures, where astrocytes have essentially no processes, oxygen-glucose deprivation (OGD) causes mitochondria to undergo a change from a highly inter-connected network to a fragmented, punctate morphology (197). In an organotypic culture in which astrocyte morphology is highly branched and more reminiscent of that observed in vivo with mitochondria found throughout the processes (for discussion see 198), transient OGD causes relatively rapid fragmentation of mitochondria in the processes followed by gradual decrease in the occupancy of the process by mitochondria (199). Similar effects have been observed in an in vivo model of global ischemia (183, 200). In preliminary studies, our laboratory has observed similar effects as early as 1 hour after photothrombosis-induced occlusion of the middle cerebral artery in mice (Shih et al., unpublished observations, Fig. 1).

The loss of astrocytic mitochondria has been linked to two mechanisms. First, damaged mitochondria (those that have been subjected to increased reactive oxygen species) undergo a specialized form of degradation called mitophagy (for reviews, see 201, 202). Quintana and colleagues showed that hypoxia and post-hypoxia reoxygenation of rat primary astrocytes results in a fragmentation of mitochondria followed by increased co-localization of mitochondria with a marker of mitophagy, microtubule-associated proteins 1A/1B light chain 3B (LC3-II) (197). This effect was associated with dephosphorylation of dynamin-related protein-1 (Drp-1) at Ser637, which results in activation of Drp1 and increased fission of mitochondria (197). Similarly, we observed an increase in co-localization of mitochondria with LC3-II after transient OGD in organotypic slices (199, Fig. 1). Second, Hayakawa and colleagues provide evidence for transfer of mitochondria from astrocytes to neurons (203). They showed that cultured astrocytes release functional mitochondria by a calcium-dependent mechanism involving CD38 and cyclic ADP ribose signaling and that mitochondria are transferred from astrocytes to neurons in culture. Furthermore, suppressing CD38 signaling with siRNA reduced extracellular mitochondria in culture, exacerbated OGD-induced neuronal damage in culture and worsened neurological damage after transient focal ischemia in vivo. These studies are consistent with the hypothesis that the loss of mitochondria from astrocytes can be attributed to both increased degradation and transfer to neighboring cells (for review, see 204). This is still a controversial observation (205), and the mechanisms that contribute to intercellular transfer of mitochondria have not been clarified (206, for review see 207).

We found that a glutamate transport inhibitor, 3-[3-[4-(trifluoromethyl)benzoylamino]benzyloxy]aspartate (TFB-TBOA), completely blocks the decrease in mitochondrial size and the loss of mitochondria observed after transient OGD in an organotypic slice (199). We also showed that glutamate, in the presence of receptor antagonists, was sufficient to cause fragmentation and loss of astrocytic mitochondria. Although one is always concerned that the effects of an inhibitor might be caused by interactions with other targets, TFB-TBOA was first synthesized in 2004, has affinities in the low nM range for blocking glutamate uptake and has no effects on glutamate receptors at 100 μM concentrations (208). Furthermore, no off-target effects have been identified to date. This suggests that either prolonged glutamate metabolism or the ionic imbalances observed with glutamate uptake contributes to the loss of mitochondria after stroke. As discussed above, activation of glutamate transport accompanied by the influx of Na⁺ activates reversed operation of the Na⁺/Ca²⁺ exchanger and increases intracellular Ca²⁺. While blocking reversed Na⁺/Ca²⁺ exchange completely attenuates OGD-induced loss of neurons, it only partially attenuates the fragmentation and loss of astrocytic mitochondria (199).

The functional consequences of this loss of mitochondria have not been systematically examined. When mitochondria undergo a change in shape from long tubular structures to small rounded ball-like structures, this increases the ratio of mitochondrial surface area to matrix, presumably increasing mitochondrial efficiency (For review, see 209). Fragmentation is also a prerequisite for degradation of mitochondria (202). In addition, fragmentation of mitochondria may be required to enable the transfer of mitochondria from astrocytes to other cells (Fig. 1, see above). Although the technology to selectively tag and isolate astrocytic mitochondria has recently been developed (210), nobody has yet used this approach to perform an analysis of astrocytic mitochondrial function after an in vivo stroke. However, several groups have taken advantage of the fact that the mitochondrial enzyme pyruvate carboxylase is selectively expressed in astrocytes in the mammalian brain (for reviews, see 211, 212, 213) to determine if stroke has a functional consequence for astrocytic mitochondria. In several different models and in both adult and neonatal animals, stroke or hypoxia/ischemia causes a decrease in flux through the pyruvate carboxylase pathway in vivo (214–216). It does appear that increasing glutamate oxidation can be beneficial within the context of stroke. For example, mice that over-express glutamate dehydrogenase have smaller lesions and reduced edema following ischemia-reperfusion (217). Similarly, a GDH activator increases oxidative glutamate metabolism and enhances neuronal survival in an in vivo model of stroke (218). Finally, mitochondria are known to provide a sink for cytosolic Ca²⁺, essentially shaping Ca²⁺ responses by decreasing amplitude of the responses and/or accelerating the decay of these signals (for review, see 219). In astrocytes, mitochondria also generate small transient Ca²⁺ elevations (111). In organotypic slices, we found that the lateral spread of spontaneous Ca²⁺ signals was bounded by mitochondria and that interfering with mitochondria or their docking increased the amplitude and spread of these Ca²⁺ signals (66, 199). We also found that transient OGD resulted in a dramatic increase in the size and lateral spread of spontaneous Ca²⁺ signals (199). Similar increases in Ca²⁺ signals in astrocyte cell bodies have been observed in an in vivo model of stroke (for review, see 220).

Conclusions/summarize future directions

Over the last forty years, glutamate has gained acceptance as the predominant excitatory neurotransmitter in the mammalian nervous system. It seems remarkably efficient that the predominant inhibitory neurotransmitter GABA is only one metabolic step from glutamate. It is clear that glutamate transporters contribute to signaling within astrocytes, between astrocytes, and in other cell types. Na⁺ influx contributes to metabolic signaling and could indirectly change the concentrations of other ions such as Ca²⁺. Ca²⁺ signaling in turn plays various signaling roles in astrocytes, including metabolism, neurovascular coupling, morphological signaling, and release of gliotransmitters. Finally, the chloride conductance that has been observed in these transporters is also used for signaling. While this has mostly been studied in the presynaptic EAATs, there is some evidence for a developmental role of these Cl⁻ currents in glial cells. Most of these findings have been in cultured cells in vitro or in acute slices, and the contributions of glutamate transport-dependent signals to in vivo astrocyte biology have not been defined.

It is also clear that the astrocytic glutamate transporters co-compartmentalize with mitochondria and that a variable amount of glutamate is oxidized in mitochondria. It seems like there are redundant (or complementary) enzymes involved in the oxidation of glutamate. However, there are essentially no selective pharmacological agents that can be used to explore the functional consequences of acute inhibition of any of these transporters/enzymes, and genetic deletion of many of the key enzymes is likely to trigger homeostatic compensation that will make it difficult to define functional roles. As indicated above, there is some evidence that the higher levels of glutamate results in greater fractional oxidation of glutamate in astrocyte cultures. One wonders if this occurs during normal synaptic transmission or if pathologic elevations of extracellular glutamate are required. The ability to generate genetically-encoded optical sensors for virtually any molecule is likely to dramatically impact our understanding of how these various biochemical pathways operate in vivo.

Finally, stroke or stroke-like insults cause rapid changes in the morphology of astrocytic mitochondria, and there is evidence that this is associated with a deficiency in anaplerotic re-generation of the carbon backbone required for synthesis of glutamate as well as exaggerated astrocytic Ca²⁺ signaling. Whether this process contributes to the pathologic progression of stroke-induced damage or is a protective response to the insult is un-resolved.